I sure hope terrorists don't like sunglasses

I think the subject of face-recognition technology, which Chris called attention to in his post about biometrics, deserves consideration of the viewpoint of the independent press, and not just the military, though I am grateful to Chris for calling attention to this subject with that post. This 2002 Wired Magazine article chronicles some of the (extremely problematic) flaws of such a technology.

Airports have been calling for face-recognition technology to scan for terrorists and other suspected bad guys in security lines since September 11 and even before. But what seems like the golden bullet for aviation security is not yet anything to get excited about, because it doesn't work.

You'll notice that while the Department of Homeland Security has a budget of $46 billion for FY 2008 (see the official DHS budget) and spends a great chunk of that on airport security, you haven't seen facial-recognition technology adopted. It's been tested many times, but it has failed.

At Palm Beach International Airport, where the technology was being tested, the computer failed to recognize airport employees 53 percent of the time. How could the computer possibly pick out a terrorist, even one whose face was known to the NSA (a hopeful assumption), at a checkpoint that millions of non-terrorists pass through every year? Right now, it's conceivable that the technology would sound the alarm dozens of times daily because grandpa has similar eyebrows to Osama bin Laden while almost certainly missing an actual terrorist.

(Nevermind the very non-trivial concerns of the American Civil Liberties Union that we ought to think twice about whether we want the government to do something like this.)

Internet privacy and security consultant Richard Smith, based in Brookline, Mass., analyzed whether this technology was "ready for prime time." His findings (In short, "no.") should be sobering to anyone who thinks this technology is anywhere close to being so.

The technology works by converting a continuous video stream of faces and converting it to digital stills. The computer then compares the faces with a database, sounding the alarm if a close match is made. The problem? The system is incredibly fussy. Sunglasses--even regular glasses--can completely throw the system, which uses eyes as anchor points.

The setup that would need to be implemented at an airport security checkpoint for the system to have any chance at working, according to Smith, would be a special walkway where lighting is "strictly controlled," with cameras at head height and passengers looking straight ahead, having removed their sunglasses and hats. That's not so hard to imagine (it hardly sounds that much more inconvenient than removing one's shoes), and DHS would certainly pay for the upgrade, which would require all new security cameras, among other things. But even if this technology were perfected a few years down the road and such a "walkway" system were implemented nationwide, there remains the problem of matching these faces to terrorists.

The real NSA is not like the Counter-Terrorist Unit of Fox's extremely right-wing (though very cool) television show "24," which seems to have every imaginable technological absurdity at its disposal. As Smith points out, it's not easy to put together a "terrorist database." Some reasons why:
1. We don't usually know who the terrorists are (Only 2 of the 19 Sept. 11 hijackers, Nawaf Alhazmi and Khalid Almihdhar, were known to the CIA and FBI, and the agencies only had a photo of the latter.)
2. Terrorists don't participate in photo line-ups.
3. Both Alhazmi and Almihdhar has US Visas and had entered the US legally twice before September 11th.

But Smith's best reason why a technology like this is a waste of time and money in trying to find even the limited number of known terrorists like Alhazmi, those who have portrait-quality photos available on MySpace and are courteous enough not to wear a hat indoors, is this: you can find Nasaf Alhazmi in the White Pages.